Anda di halaman 1dari 11

The Nuclear Decay of Cesium-137 Andrew Mowry with Nick Bern

March 9, 2012

Abstract We present in this paper some evidence for that basic assertion of quantum physics, that at the particle level interactions are governed by probability. We frame this topic in the context of nuclear decay, following some experiments in the counting of gamma rays as a radiation product of Cesium-137 decay. Continuing forward to an application, we also calculate the half-life of Ba-137m. Understanding the process of nuclear decay is important in todays world, if not because it is necessary in controlling nuclear processes (i.e. power generation) and deducing histories (i.e. radiometric dating), then because other people already understand it (i.e. they know how to make dangerous bombs from radioactive elements). Although the physics of nuclear reactions is not the easiest subject to wrap ones mind around, the statistics of it is quite manageable. Passing over the Whys, an unstable, radioactive compound will decay spontaneously into some number of lower energy products, a process which continues until all of the atoms in the compound have decayed. It is common knowledge that these decays cannot all occur instantaneously: otherwise there would be no great controversy over the storage of nuclear waste, for example. Rather, each unstable isotope has its own characteristic half-life, which is an amount of time over which we expect half of a samples atoms to decay. From experiments we can show the overall decay of a sample is an exponential function of the form remaining atoms = (initial atoms)eb(time


This seems to imply that there is some xed rate of decay, i.e. that each atom might have a timer attached to it since creation that portentously counts down the exact moments until nuclear reaction. This is not so. The impetus for spontaneous decay, rooted rmly in the realm of quantum physics, is merely a probability. It is a premise of this paper that for a large enough sample the probability may be relied on to deliver predictable results, but they are by no means certain. The randomness of each atoms individual decay is crucial. In this paper we take several measurements of the radiation (taken as evidence of nuclear decay) emitted from a sample of Cesium-137 that will illustrate the seemingly contradictory properties of predictability and randomness. We then proceed to calculate the half-life of Barium-137m. The relevant decay process is illustrated by Fig. 1: the nal decay product is an energetic beta particle, Ba-137, and in 94.6% of cases, specic gamma radiation.1 In this
Note that we are assuming the predictability of random processes, here that a certain percentage of Cs-137 will decay in a two step process.

2 more likely case, the Ba-137m is an intermediate particle. In any sample of Cs-137, then, we will nd that an equilibrium number of Ba-137m atoms is soon established by its creative and destructive processes, and that each one of these processes is associated with its own radiation. The mean number of nuclear decays in any given time period should be static, then, given a large enough sample size.

Figure 1: The nuclear decay schedule for Cs-137. [3] We begin by setting a button source of Cs-137 5n front of our NaI crystal, nearly touching them together.2 3 Setting the NI-DAQ to count 1024 100ms bins, we collect the data shown in Fig. 2. We also overlay the Poisson and Gaussian distributions for this sample space, using the observed mean count of 89.678.4 Calculating the standard deviation of our data set directly, we see that =
N i=1 (ni

)2 = N 1

1024 i=1 (ni

89.678)2 = 9.5734 1023


and the standard deviation p as calculated from the Poisson/Gaussian distribution yields p = = 9.4698. (3) The very small dierence between these numbers (a 0.1% dierence relative to the mean value) is evidence that the underlying mechanism of nuclear decay is random. We reach this conclusion because the collected data correlates with the Gaussian models that take as their premises complete randomness. Evidently, the bars of the histogram in Fig. 2 do not align perfectly with either the Poisson curve or the Gaussian curve. We restate clearly that this is because individual
See Appendix .1 for details on the apparatus The particulars of the source are not relevant, as we are only looking for a qualitative behavior to describe. 4 See Appendix .2 for a short discussion of these two distributions.
3 2

3 nuclear decay events are fundamentally random. As we add more and more data, we expect the average of it all to mold to the theoretical curve, but we still anticipate slight deviations.5 It is also barely perceptible from Fig. 2 that the Poisson and Gaussian distributions are not identical, but correspond closely. We keep this in mind moving forward.

Figure 2: Empirical and theoretical distributions for Cs-137 decay, with the source placed close to the scintillation detector. 1024 100ms bins. Our second test maintains the same conditions of the rst test, but we increase the size of each of the 1024 bins to 1 second from .1 seconds. Intuitively we would expect ten times as many events per bin, and this is represented in the mean, = 898.075. A graphical representation of the results is presented in Fig. 3, along with the Gaussian distribution curve.6 From the picture it is easy to see that the empirical distribution follows the Gaussian distribution curve on average. We calculate as above = 29.043 and p = 29.968,


Taking an ensemble average with 15 systems, for example, should produce a more perfectly-shaped histogram than an ensemble average of 5 systems, which should produce a nicer histogram than our 1 system. 6 We omit the Poisson distribution curve, although it would be a welcome addition, because it is computationally unfeasible to generate for such large mean counts. (The computer must maintain in memory the factorial of 898, and e raised to that power.)

4 which again deviate from each other by only 0.1% relative to the mean value, i.e. there is a very strong correlation between the data and the Gaussian distribution. This implies, once again, that the nuclear decay process is fundamentally random.

Figure 3: Empirical and theoretical distributions for Cs-137 decay, with the source placed close to the scintillation detector. 1024 1s bins. Finally, we consider the data in Fig. 4 from the Cesium-137 source placed behind a lead block, which is a very good radiation shield against the particular decay products. The counts were captured in 1024 50 ms bins. We see some interesting behavior here in that the Poisson and Gaussian distributions do not correspond very well. This is expected, however, from our prediction in Appendix .2, because the average count, = 0.64258, is very small, and so the Poisson distribution is not a valid approximation of the Gaussian distribution. The good correspondence between the two models in the rst test was so precisely because was much larger. Again calculating the standard deviations, we nd that = 0.80476 and p = 0.80161. The similarity of these results is just one more piece of evidence that the nuclear decay process is characterized by randomness.

Figure 4: Empirical and theoretical distributions for Cs-137, placed behind a lead block. 1024 50ms bins.

6 The half-life of Ba-137m We move now to a more quantitative exercise, having seen that the behavior of nuclear decay is well characterized as a random yet predictable process. The distributions that we were able to make were importantly dependent on the equilibrium between Cs-137 decay and Ba-137m decay. This ensured that expected number of radiation events per any given time remained constant, i.e. did not change as we ran the counting program. If we alter our source, however, to break equilibrium, the counting program should give us data indicating the rate of decay and half-life of the sample. This is what we do, using a proprietary elution process to separate a population of Ba-137m from their parent Cesium population. Dissolving this unknown population of Ba-137m into about 1 mL of water, we set the sample an inch away from the scintillation detector. We instruct the program to count radiation for 1024 1 second bins, and it returns the results in Fig. 5.

Figure 5: The empirical and theoretical distributions for Ba-137m with 1024 1 sec bins. The uncertainties (1st standard deviations) are weighted by the inverse of the standard deviation squared.

It is immediately apparent to the eye that the set of points (each of which records a number of radiation events in a given time bin) takes on a roughly exponential form. This is not a surprising result, as exponential decay has the properties that we expect in nuclear decay: a xed half-life and a monotonically decreasing behavior (on the average the randomness of decay means slight deviations are not problematic).

7 We use SigmaPlot to t a three parameter exponential decay function to our data, and get a very good result: the function f (t) = 12.77 + 2860.7 e(.0045t) , (5)

which is shown on top of the data in Fig. 5, has an R2 value of .9978 an exceedingly good t. The value (.0045), usually denoted , is the nuclear decay coecient and describes the rate of decay. The value (2860.7) is approximately equal to the initial population of radioactive particles, and the value (12.77) is an oset. This last value is a small issue, because we would expect an oset of 0. After all, as time stretches to innity, we expect the counts to approach 0. We posit that there may be some constant background radiation triggering the detector and osetting our data. To test this, we remove our source Ba-137m from the vicinity of our apparatus as best as possible, and initiate the program.7 Using the customary 1024 bins set at 50 ms per bin, we report a value of .626(786) counts per 50 ms, or 12.36(352) counts per second. This matches the oset value of the decay data reasonably well, and we are satised that we have thus identied and quantied the culprit. Compensating our initial data set with this new information (and propagating uncertainties) leads us to replot and ret our data as shown in Fig. 6. The new function of best t is f (t) = .736 + 2860.4 e(.0045t) , (6) which has an R2 value of .9982: another fantastic t.8 We are now fully equipped to calculate the half-life of Ba-137m. Starting from the functional form of exponential decay with a starting population A:9 1 ln(2) A = Aet1/2 ln( ) = t1/2 . (7) 2 2 The half life is thus simple to calculate. Using the decay coecient from our corrected t, we determine that the half-life t1/2 of Ba-137m is 154.0(4) seconds. This compares favorably (within two standard deviations) with the accepted value: 153.12(6) seconds[1]. We have then completed the two purposes of this paper, to show that nuclear decay, while random on the particle level, is predictable in large systems; and that accurate measurements of half-life are possible based on that premise. To return to the motivation of the paper, that widespread knowledge of radioactive decay is imperative in modern society, we have uncovered both the fundamental principle and a basic calculation of the topic. The ways to advance from this point of knowledge are two: to delve deeper into the premise itself and question why or if, in fact, the process is random; or to follow the consequences of that f (t) = Ae(t)
Eliminating the source completely seemed slightly challenging, but we removed it to another table and stued soft tissue into the solution. 8 Note that the nal oset of .736 actually has an uncertainty of .818 attached to it, so the theoretically correct 0 oset is plausible. 9 We omit the oset value because it has an inconsequential eect on the half-life, having no eect on the actual rate of decay.

Figure 6: The empirical and theoretical distributions for Ba-137m with 1024 1 sec bins. The data has been corrected for background radiation. The uncertainties (1st standard deviations) are weighted by the inverse of the standard deviation squared.

statement into more advanced and useful calculation. Undoubtedly the paths both need to be tread to advance human understanding of nature.



We use the system of devices shown in Fig. 7 to collect data for the Cesium-137 and Barium137m decay testing. The sodium iodine crystal is a scintillation detector, which uoresces in response to incident ionizing radiation (i.e. alpha, beta, and gamma radiation). The photons produced by this uorescence are directed into the photomultiplier tube, where an electrical signal is created to communicate with the rest of the apparatus. Because the nuclear decay radiation is isotropic, the rate of radiation detection is proportional to the actual rate of decay. After the signal passes through the preamp and spectroscopy amplier, it enters the Single Channel Analyzer (SCA) in a manipulatable state. Through the SCA interface, we are able to lter the incoming signals and standardize any output. We do this, choosing only to accept signals with an amplitude between .5 V and 7.5 V. This setting removes any processing noise or high energy events (cosmic ray interaction) from inuencing the nal counting. The NI-DAQ is the actual signal pulse counter, and is controlled by a LabVIEW program that sets the number and width of bins, which are the periods that the NI-DAQ counts pulses for. Data is ultimately collected by the computer for analysis.

Figure 7: The system used for data acquisition. [2]


Probability Distributions

We make use of two probability distributions in this paper: the Gaussian and the Poisson. Each is a function Pk that determines the normalized probability of measuring k independent events for an expected value .10 The distributions are

The expected value, or mean value, can be deduced experimentally as we do here, or theoretically.


k e e 2 , and Poisson : Pk = . Gaussian : Pk = 2 k!



The Poisson distribution is actually a limit of the Gaussian distribution for large , which makes it applicable for many situations such as radioactive decay (although we must be wary to avoid misuse, as in Fig. 4). Besides oering a more elegant function, the Poisson distribution has the nice property that the standard deviation p is exactly 1/2 . It is attractive because of its simplicity.

[1] Audi, G., Bersillon, O., Blachot, J., & Wapstra, A. (2003). The nubase evaluation of nuclear and decay properties. Sponsored by CSNSM, SPN, NIKHEF. Retrieved from on 3/8/2012. [2] Hall, David. (2012). Labs and Lectures, Information and Figures. Spring 2012. [3] Nave, C.R. (2012). Hyperphysics, Georgia State University. Cesium-137. Retrieved from on 3/9/2012.